Connect Claude with MCP

Notebooker ships a remote MCP (Model Context Protocol) server, so AI assistants that support MCP connectors - Claude on the web, desktop, and mobile among them - can work with your library directly: list your notebooks, add sources, write notes, search everything, and ask questions across your research.

There’s nothing to install and no API key to copy. The connector uses OAuth: you sign in to Notebooker once, approve access, and the assistant gets its own revocable credentials for your account.

Connecting from Claude

1. In Claude, open Settings → Connectors and choose Add custom connector.

2. Enter the connector URL:

   https://app.notebooker.ai/mcp
   

3. Claude opens a Notebooker sign-in window. Log in (if you aren’t already) and click Authorize on the consent screen.
4. That’s it - Claude can now see the Notebooker tools. Try “list my notebooks” or “add this article to my Reading notebook”.

Any other MCP client that supports remote servers with OAuth works the same way - point it at the same URL and complete the sign-in flow.

To disconnect, remove the connector in your client’s settings. The client’s access tokens expire on their own within a day and refresh only while the connector stays installed.

What the connector can do

Tool	What it does
manage-notebooks	List, create, rename, archive, or delete notebooks.
manage-sources	List sources, add links or pasted text, check processing status, retitle, retry, or delete.
manage-notes	List, read, write, update, or delete notes.
search	Keyword or semantic search across your sources and notes.
ask	Ask a question across your whole library and get a cited AI answer.
chat	Chat with a specific notebook’s sources, with retrieval.
transformations	List your AI transformations or run one on text.
insights	Generate, read, and save AI insights on sources.

A few things worth knowing:

• Adding sources is asynchronous. When the assistant adds a link, Notebooker fetches, extracts, and embeds it in the background - the same pipeline as adding a source in the app. The assistant gets a status tool to poll until processing finishes.
• Everything is scoped to your account. The connector sees exactly the notebooks, sources, and notes your login sees - nothing else.
• AI actions draw on your credit. Asking questions, chatting, running transformations, and semantic search use your monthly AI credit exactly like the same actions in the app - see Usage, credit & budgets. Reading and searching by keyword are free, and reads keep working even when your credit runs out.
• Deletes are real. If you ask the assistant to delete a notebook or source, it’s deleted - same as doing it in the app. Claude asks you to confirm tool calls that change data, so read those prompts before approving.

MCP or the API?

Use the MCP connector when an AI assistant should work with your library conversationally. If you want to drive Notebooker from scripts, n8n, or anything that speaks plain HTTP, personal API keys are the better fit.